PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Towards Intent Dependent Image Enhancement: State-of-the-art and Recent Attempts
Marco Bressan, Gabriela Csurka and Sebastian Favre
In: VISAPP 2007, March 2007, Barcelona, Spain.

Abstract

Image enhancement is mostly driven by intent and its future largely relies on our ability to map the space of intentions with the space of possible enhancements. Taking into account the semantic content of an image is an important step in this direction where contextual and aesthetic dimensions are also likely to have an important role. In this article we detail the state-of-the-art and some recent efforts in for semantic or content-dependent enhancement. Through a concrete example we also show how image understanding and image enhancement tools can be brought together. We show how the mapping between semantic space and enhancements can be learnt from user evaluations when the purpose is subjective quality measured by user preference. This is done by introducing a discretization of both spaces and notions of coherence, agreement and relevance to the user response. Another example illustrates the feasibility of solving the situation where the binary option of whether or not to enhance is considered.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Machine Vision
ID Code:3488
Deposited By:Gabriela Csurka
Deposited On:11 February 2008